The most important number attached to a proposed data center may no longer be its megawatt capacity. It may be the value created by each megawatt once the facility starts operating. That distinction matters because electricity is becoming an increasingly strategic input for digital infrastructure. The International Energy Agency expects global data center electricity consumption to roughly double from 485 TWh in 2025 to about 950 TWh in 2030, while AI-focused facilities could see even faster growth. At the same time, the agency says electricity consumption per AI task is falling rapidly as computing efficiency improves.
The contradiction is becoming harder to ignore. Computing is getting more efficient, but the amount of computing society wants is expanding even faster. That leaves an uncomfortable question for developers, utilities and investors: when electricity supply remains constrained, should every credible data center proposal receive the same claim on scarce power? The answer cannot simply rest on whether a project has secured land, financing or a connection application. A power request is not the same thing as economic value. A permitted megawatt is not automatically a productive megawatt.
A megawatt should have to prove what it delivers
The traditional infrastructure argument treats capacity as the achievement. If a developer can assemble land, secure a connection and announce a large campus, the project begins to look like progress before it has delivered a single computational workload. That logic becomes weaker when several projects compete for the same electricity. A data center consumes power continuously to provide computing, storage, networking and increasingly AI workloads. Those services can generate substantial economic activity, but the relationship between electricity consumed and value produced varies enormously between facilities. A high-utilisation AI cluster serving commercially valuable workloads is not economically identical to a facility whose capacity sits idle while its operator waits for customers.
The relevant metric is not simply megawatts secured. It is productive output per unit of electricity, adjusted for utilisation, computing efficiency, water consumption, economic contribution and the additional clean generation required to support the load. Such a framework would not require policymakers to decide which companies deserve to operate. It would require projects to demonstrate what they intend to do with the electricity they request.
Spain offers a useful warning about capacity and demand
Spain illustrates why the distinction matters. The country has emerged as an attractive location for data center investment because of its renewable power resources, land availability and growing digital infrastructure market. Yet the scale of proposed projects creates a difficult planning question. The IEA found that planned data center projects in Spain are seven times larger on average than facilities currently operating. If the European project pipeline were fully realised, data centers could represent about 10% of current peak electricity demand in Spain.
Spain’s electricity planning process also shows the scale of the wider demand queue. The government’s 2030 planning proposal identifies 27.7 GW of demand connected to the transmission network, alongside 5.3 GW of support for distribution networks. Officials have explicitly linked the planning approach to the need to distinguish genuine demand from projects that may never materialise. That does not mean Spain has 27.7 GW of confirmed data center consumption. It does not. The figure covers broader demand, including industrial projects and other large loads. A connection request represents an intention to consume electricity. It does not establish that the electricity will produce the projected economic or technological output.
Clean power should mean additional power, not accounting
The same principle applies to renewable energy. A data center can contract renewable electricity and still raise difficult questions about whether its consumption creates additional clean generation or simply competes for existing supply. Spain’s expanding renewable generation gives it an important advantage, but renewable electricity remains a physical system rather than an accounting category. If large new loads arrive faster than new generation and network capacity, the system still has to balance them.
That makes additionality increasingly important. New data center demand should encourage new generation, storage, transmission capacity and flexibility where possible. A project that helps finance additional clean capacity creates a different system effect from one that simply claims existing renewable output. Water efficiency belongs in the same calculation. Cooling technology can materially change the resource intensity of a facility, particularly in regions where electricity and water constraints overlap. A project that delivers substantially more computing while reducing both electrical and water intensity should have a fundamentally different resource profile from an inefficient facility delivering the same digital output.
The unit of competition is changing
The data center industry has become accustomed to thinking in terms of scale. Bigger campuses promise more capacity, more customers and greater investment. But electricity scarcity changes the equation. When power becomes the limiting input, scale alone stops being an adequate measure of progress. A 500 MW facility cannot claim greater social or technological value simply because it consumes five times the electricity of a 100 MW facility. The more consequential question is what those megawatts accomplish.
That shifts the debate from bytes versus watts to value per watt. It asks whether computing demand is real, whether infrastructure operates efficiently, whether water use remains proportionate, whether new clean generation follows new consumption and whether the resulting digital capability justifies the electricity allocation. The next data center race may therefore produce a less impressive headline but a more useful metric. The industry will still count megawatts. Investors will still announce campuses. Developers will still compete for power. But the projects that matter most will increasingly have to explain what society receives in return for every watt they consume. That is a much harder proposition than securing a connection. It is also the question that a power-constrained digital economy can no longer afford to avoid.
